We have identified two effects that backward compatibility for, say, agents of y type has on two-sided market (the argument is naturally the same for the backward com-patibility for agents of x type). First effect is a direct effect that is positive for new y-agents (backward compatibility improves their incentives) and is negative for old x-agents (backward compatibility discourages them from buying the new technology).
Second effect is a negative feedback effect. Namely, decrease in the demand of old agents of type x leads to the decrease in the demand of old agents of type y and to the decrease in the demand of new agents of type y. The negative feedback effects
become more important if the technological progress is revolutionary (s and B are large), while the direct positive effect less so. Indeed, if the new platform is very advanced, then the new agents have large incentives to buy it even if it does not allow them to access the installed base of agents on the other side of the market. The compatibility therefore will bring only moderate improvement in their demand.
The tradeoff between direct effect and feedback effects determines the optimal compat-ibility choice. In particular, as follows from our analysis, the backward compatcompat-ibility is more likely to be imposed on the market where the technological progress is mod-erate.18 Further (other things equal), the compatibility for agents of type y is more likely to be imposed if their installed base is relatively small, the growth rate of the installed base is moderate and the installed base ofx-agents is small.
Our model provides several predictions about patterns of compatibility choice. On the emerging market, where technological progress is rapid and the entry of agents on both sides of the market is significant, we should often observe the subsequent generations of technologies being not compatible with each other. As, however, the pace of technological improvement slows down and the growth of installed base decel-erates (the market becomes mature), we should expect some degree of compatibility between subsequent generations of technology. In particular, technologies are likely to be backward compatible for some side of the market if the installed base on this side of the market is relatively small. Technologies are likely to be fully compatible if the both sides of the market are symmetrically represented. Only if the technological progress is significant and the installed base on both sides of the market are very large should the technologies remain incompatible.
The predicted pattern of the compatibility regimes as a market develops from emerging to mature is nicely illustrated by the experience of video game console industry. The following discussion is adopted mainly from the Evans et al. (2006). The start of the video game console industry dates back to the earlier 70’s. However, the industry was emerging at the slow pace. The leader of the industry, Atari, at the pick of it success sold only around 5 mln units of video game consoles. Moreover, the game industry crashed in 1983 due to the overproduction of poor quality games. The credit for the revival of the industry goes to Nintendo. Around 1983 Nintendo introduced its first console (Nintendo Entertainment System, NES) which has revolutionized the way how the video console business was done. Nintendo actively pursued a two-sided
18Note, that this argument does not rely on the costs of achieving compatibility, which are nat-urally higher if the platforms belong to the very distant generations of technology. Higher costs provide another reasons why the platforms should be incompatible if the technological progress is revolutionary.
market strategy. It drafted licensing agreements with third party providers to ensure the quality of the games and the critical mass of the games for the new system. The sales of the NES and related games skyrocketed. It sold around 60 mln consoles world wide.
Ninetendo operated at a clearly emerging market, where the pace of technological growth was rapid and the installed base of users of consoles and game developers was relatively small. The future generations of Nintendo video game consoles were incom-patible with the previous version.19 Super NES (introduced in 1990) was incompatible with its predecessor NES; Nintendo 64 (introduced in 1996) was incompatible with Super NES and Game Cube (introduces in 2001) was incompatible with Nintendo 64.
Presently, in Japan, USA and Europe, the video game console market has reached its mature state. According to estimation of Nielsen, a market research company, 41.1%
of US households own a game console and the rate of console penetration has slowed down20. In line with our predictions Nintendo made its new console, Wii (introduced in 2006) backward compatible with its predeccessor, Game Cube.
Our analysis also indicates how the compatibility of platforms should evolve on the asymmetric market where a monopolist, who previously treated his market as a one-sided business decides to disintegrate and to embrace a two-one-sided model. Following Section 3.4.3 let us assume that there is an installed base only on the x side of the market and the growth rate of the market are cx and cy. Then we would expect subsequent generations of platforms be compatible for agents of typey, ifcy is small and/or if the pace of technological progress is moderate.
To illustrate this prediction, consider the case of Palm company.21 Palm started as a software company but soon integrated in a hardware. Although it did not produced the hardware itself, it controlled all stages of the process and treated the involved firms as subcontractors. PalmPilot, produced in 1996, was a hardware with inte-grated operation system. However, in late 1997 Palm switched to a two-sided model.
It has concentrated on the development of Palm OS operation system, which it was licensing to the hardware makers, such as Sony, Kyocera, Nokia, Handspring etc. (y agents, in the terminology of our model). On the other hand, to ensure the popularity
19Interesting enough, there are add-ons, unlicensed by Ninetendo, which allow to make subsequent generations of Nintendo consoles compatible. It indicates that Nintendo decided to make platforms incompatible not because it was technically impossible, but because it was more profitable strategy. The information about backward compatibility is taken from en.wikipedia.org/wiki/Backward compatibility.
20See report of The Nielsen Company (2006)
21The example is taken from Evans et al. (2006, Ch. 6 and Ch. 9). The data about compatibility of Palm OS is available atwww.access-company.com/developers/documents/docs/palmos.
of Palm OS, Palm has intensively courted the developers of applications (the x side of the market) from the time of introduction of Palm Pilot. It already had significant installed base of third party developers when it decided to switch to the two-sided model. The efficient courting strategy ensured that this base was growing fast. How-ever, due to some management failures and the market trends, Palm had less success in ensuring the cooperation of third party providers of hardware. Sony, for example, has stopped selling PDA’s which run Palm OS. Handspring was purchased by Pal-mOne (a hardware company, independent from PalmSource, provider of Palm OS). In line with our prediction, Palm, eager to improve attractiveness of its operation system for the hardware developers, made it backward compatible. Any version of Palm OS, installed on hardware device, is not only able to run the applications, written for this version but also applications written for the older versions of the operation system.